elder-plinius/CL4R1T4S

The repository aggregates extracted system prompts, behavioral guidelines, and tool definitions from major AI providers (OpenAI, Google, Anthropic, xAI, Perplexity, and specialized coding agents), making visible the hidden instruction layers that shape model outputs and constrain behavior. It operates as a crowdsourced archive where contributors submit reverse-engineered or leaked prompts with metadata (extraction date, context), enabling comparative analysis of how different vendors implement safety constraints, persona enforcement, and content policies. The project frames system prompt transparency as essential for understanding AI systems as non-neutral intelligence tools rather than objective information sources.

13,738 stars.